Path Dependence in European Development:

Medieval Politics, Conflict and State Building ∗

Avidit Acharya† Alexander Lee‡

December 13, 2018

Abstract

During the Middle Ages, most European polities operated under a norm that gaveonly the close male relatives of a monarch a privileged place in the order of succession.When no such heirs were available, succession disputes were more likely, with distantrelatives and female(-line) heirs laying competing claims to the throne. These disputesoften produced conflicts that destroyed existing institutions and harmed subsequenteconomic development. A shortage of male heirs to a European monarchy in the MiddleAges could thus have harmful effects on the development trajectories of regions ruledby that monarchy. We provide evidence for this by showing that regions that weremore likely to have a shortage of male heirs are today poorer than other regions. Ourfinding highlights the importance of the medieval period in European development andshows how small shocks can work in combination with institutions and norms in shapinglong-run development paths.

Key words: political instability, conflict, state-building, economic development, gen-der bias, political norms and institutions, European history

∗We are grateful to Lisa Blaydes, Gary Cox, Edo Grillo, Jens Hainmueller, Andrej Kokkonen, DavidStasavage, and seminar audiences at the University of Rochester, Juan March, LSE, MPSA, Nottingham,Oxford, UC Berkeley, UCLA, and Yale for valuable comments and suggestions. Riva Yeo, Julie Dib, JillianJohannes, Madalina Ciocanu, and especially Scott Williamson provided outstanding research assistance.†Assistant Professor of Political Science, Stanford University, Encina Hall West, Stanford, CA 94305.

Email: avidit@stanford.edu.‡Assistant Professor of Political Science, University of Rochester, Harkness Hall, Rochester, NY 14627.

Email: alexander.mark.lee@rochester.edu.

1

1 Introduction

A large body of scholarship argues that the process of political and economic development ispath dependent. Most works in this “historical institutionalism” literature focus on studyingthe consequences of large, often deliberately initiated, macro-historical events that took placeat certain “critical junctures” in shaping long run development trajectories.1 Some of thiswork, however, also points to the possibility that even small contingent events can producepath dependent outcomes. Paul Pierson, for example, writes that “path dependent argumentsbased on positive feedback suggest that not only ‘big’ events have big consequences; littleones that happen at the right time can have major consequences as well” (Pierson, 2000,p. 263). Despite this, empirical studies of the large, long term consequences of such smallcontingent events have been relatively scarce. This paper contributes one such study.

Our focus in this paper is on the development of state institutions in medieval Europe,and how accidents of royal births—in particular, the availability of male heirs to the monarchsof medieval Europe—contributed to producing a variation in the success or failure of earlystate-building efforts. The effects of this variation appear to have lasted and can be seenin contemporary differences in development levels across Europe, overlaying other, betterknown dimensions of economic variation within the continent.

The link between royal births and medieval state development emerges from a simpletheory connecting political gender bias and succession norms to political conflict and itsconsequences for state-building. During the medieval period, most polities operated undera norm that gave the close male relatives of the ruling monarch a relatively clear place inthe line of succession, but often did not specify clear succession rules for female (and female-line) heirs. A shortage of male heirs could, therefore, destabilize the polity by creatingconflict among the nobility, rival relatives and opportunistic foreigners. In many cases, moredistant male relatives would claim the throne over closer female or female-line relatives,leading to a succession dispute. These disputes often turned into violent wars. These warsin turn had harmful effects on the polity’s long term development, by destroying existingstate institutions and reducing political cohesion. It is also possible that ruling families andlocal elites in areas ruled by monarchs with relatively few close male relatives anticipatedsuch conflict, and became less willing to invest in building a strong and cohesive state inthe first place. Both this underinvestment, and the damage caused by violent conflict had adetrimental effect on contemporaneous and subsequent state development.

1See, e.g., Lipset and Rokkan (1967), North (1990), Skocpol (1979) and Thelen (2004), and Collier andCollier (2002), among many others.

1

In areas with one or more potential male heirs, on the other hand, the high level ofpolitical stability afforded by a series of unambiguous and uncontested successions enabledcertain polities to establish strong and lasting state institutions, making them more resistantto internal conflict and foreign influence, and even enabling them to conquer and exploit otherless politically stable areas. Paris is one of Europe’s most prosperous metropolises today inpart because it enjoyed the uninterrupted reign of a single dynasty for more than threecenturies from 987-1316AD (Lewis, 1981). It was in this period that Capetian monarchslaid the foundations of the modern French state with their city as its capital. Naples, bycontrast, was a more prosperous city than Paris in the early Middle Ages, but lost grounddue to a series of destructive civil wars brought on by succession-related conflicts (Jones,2000). Our theory provides a new explanation for this reversal.

Our main empirical findings demonstrate the path dependent effects of the uneven natureof state-building in medieval Europe arising from the (un)availability of male heirs. Regionsof Europe that were ruled by medieval monarchs who had an abundance of male heirs aretoday richer than other regions. These effects can also be traced over time: urban densityin each century between 1300 and 1800 was higher in regions that had an abundance ofmale heirs. In addition, we find that an abundance of male heirs decreased the frequencyof internal wars and coups, and that contemporary economic development is negativelycorrelated with the occurrence of these violent medieval events, lending further support toour theory. To allay some concerns about endogeneity, we also conduct a placebo test thatshows that availability of illegitimate male children (those who did not have claims to succeedthe throne) does not predict prosperity today.

Authors such as Strayer (1970) and Tilly (1992) have argued that European polities beganto develop state institutions in the period between 1000 and 1500, but that some politiesbuilt stronger and more durable states than others. Tilly (1992) in particular highlightsthe destructive impact of internal wars and political instability on medieval state-buildingin Europe. In addition, Migdal (1988), Herbst (2000) and Besley and Persson (2011) haveargued that strong state institutions, at both the local and central level, are essential foreconomic development. Our paper provides evidence for these claims by identifying anaspect of medieval politics that is both readily measurable and has direct theoretical links tomedieval state-building efforts (the availability of male heirs), and documenting its lastingeffects on contemporary development.

Our work also relates to a vast literature that argues that societal and economic conditionspreceding the period of Western industrialization played an important role in shaping thedevelopment paths of European polities. By focusing on pre-1500 conditions and the roleof historical norms and prejudices in shaping long-run outcomes, our paper contributes to a

2

lively debate on how far back the different development trajectories of modern societies arediscernible, and what caused these divergent paths in the first place. Some scholars, such asPomeranz (2009) and Acemoglu, Johnson and Robinson (2002), for example, have arguedthat development trajectories became well-defined only in the post-1500 period, and reallyonly after the Industrial Revolution. Others, such as Maddison (2007), Putnam, Leonardiand Nanetti (1994), Guiso, Sapienza and Zingales (2016) and Chaney (2014) have arguedthat pre-1500 conditions also played an important role in setting the development trajectoriesof many European polities. While some of these authors (e.g., Acemoglu, Johnson andRobinson, 2005) highlight the persistence of institutions in explaining development paths,others emphasize the importance of culture and social norms (e.g., Guiso, Sapienza andZingales, 2016). By stressing the importance of historical norms and prejudices in shapinginstitutional paths, our paper brings together these two perspectives.2

2 Theoretical Background

We connect contemporary development to the availability of male heirs to medieval Europeanmonarchs via conflict, political instability and state-building. The theory is schematicallysummarized as follows:

shortage ofmale heirs

⇒conflict andinstability

⇒persistently weakerstate institutions

⇒persistently weaker

development outcomes

As we document below, a rich literature provides support for the second and third links above:that political instability and conflict are harmful for the development of state institutions,and strong states are important for economic development. We briefly review this literature.We then provide theoretical background to the first link in the chain above, between maleheirs and conflict.

2.1 Conflict, State Building and Development

The medieval period has long been recognized as a pivotal one for the growth of state insti-tutions. Tilly (1992) and Strayer (1970) emphasized the importance of the medieval periodfor the development of durable state institutions in Europe. In turn, Migdal (1988), Evans(1995) and Herbst (2000) argued the importance of state building for economic development.

2In this way, it is also relates to the work of Tabellini (2010), who also studies variation in developmentoutcomes across European regions, but argues that culture is the channel by which historical institutionsexplain contemporary development across European regions, rather than institutions being the channel bywhich historical norms influence modern development outcomes.

3

According to this literature, developing state institutions that are capable of providing mar-ket supporting public services, law enforcement, and protection of private property rights, isessential for laying the foundations to economic prosperity.3

The Middle Ages are a period where the link between state building and long run eco-nomic development is likely to have been especially strong, since it was during this periodthat many of the basic conditions for modern states were being created, or failing to becreated. Examples of the achievements of successful medieval states include the reductionof the ability of petty lords to levy arbitrary taxes and customs duties (North, 1973), thereduction in violence, especially the so called “private wars” within the borders of the state(Tilly, 1992), and the development of institutionalized court systems that improved legalinstitutions (Harding, 2002). These developments made investments in economic activitymore attractive by reducing the possibility of expropriation by greedy locals, bandits, andcorrupt royal officials. Protection from a large coercive monopolist with long term interestsmade obsolete many of the ad hoc techniques that medieval traders had developed to protectagainst these types of expropriation (Greif, 1993; Milgrom, North and Weingast, 1990).4

Complementary to the literature on state development, Bates (2001) argues that whilehaving a strong states is important for development, building state institutions in the pres-ence of political instability and conflict can be challenging.5 In the face of high instability,rulers know that they could be ousted at any time, and have little incentive to invest in state-building. And, in the presence of frequent conflict, their incremental investments would beperiodically destroyed anyway.6 These arguments give us a basic relationship between statecapacity and development that has been summarized and extended formally by Besley andPersson (2010). We rely on their work as well as the previous literature to justify the second

3North and Weingast (1989) and North (1981) provide a counterpoint to these claims. They cautionthat if the state is too strong, its rulers may transgress by expropriating too much from productive subjects,thereby reducing investments and economic growth. In contrast, recent work by Stasavage (2014) shows thatin many instances strong rulers were able to restrain guilds from creation barriers to entry, which would haveotherwise stifled competition and hurt development. Although no consensus exists on how strong a statemust be for it to become harmful for development, we take the view that state building in medieval Europewas particularly good for long term development because it laid the foundations for the most fundamentalinstitutions that support market activity—institutions that Migdal (1988) and Herbst (2000) have arguedare lacking in much of contemporary Africa.

4Many of the achievements of medieval states in the Early Modern period involved the removal of com-petitors to state authority rather than the creation of large bureaucracies, modern armies or changes instate capacity, but which were central to the later development of modern state institutions (Brewer, 1990;Downing, 1993)

5See Alesina and Perotti (1996) and Barro (1996) for evidence that political instability is harmful fordevelopment.

6This gives rise to the following development “trap:” although strong states are often required to limitconflict, conflict itself inhibits the development of strong states. See Cox, North and Weingast (2015) for aclosely related argument.

4

and third links of the theory represented by schematic diagram above, with two importantadditions.

First, previous work has focused on the argument that internal conflict and politicalinstability reduce investments in state capacity, in turn reducing economic investments. Inmany models of conflict and state capacity, such as that of Besley and Persson (2010), violentinternal conflict does not have any direct destructive impact on state institutions or devel-opment outcomes; it simply diverts resources away from productive activities. In additionto this, we suggest the possibility that such conflict damages existing state institutions aswell, and therefore hurts development via this second, more direct, channel.

Different types of conflicts might have different effects. Besley and Persson (2010) makethe distinction between internal conflicts (civil wars) and external conflicts (interstate wars),and argue that internal conflicts hurt state building while external conflicts promote statebuilding. The importance of this distinction was first suggested by authors in the bellicisttradition such as Tilly (1992). In other recent work, Dincecco and Onorato (2013) andVoigtländer and Voth (2013) find a positive relationship between external war and urban-ization in the early modern period, which they trace to the statebuilding generated by thedemands of external conflict. For our purposes, the relevant wars are internal wars, becauseconflicts emerging from succession disputes are more likely to be internal (though some im-portant wars emerging from succession disputes, such as the Hundred Years War, count asboth internal and external, depending on the country involved).

Second, most previous work on state capacity and development provides only a proximatetheory of the impact of state capacity on development. Nevertheless, differences in state de-velopment that arose in the distant past are likely to have persisted, making it possible in ourcontext that European polities in which conflict destroyed or stunted the development of thestate in the medieval period have persistently lagged behind those in which such conflict didnot destroy state institutions as severely. This view is consistent with an influential literaturethat argues that historical institutions are generally very persistent, and their developmentover time tends to be incremental (North, 1990; Acemoglu, Johnson and Robinson, 2005).

2.2 Male Heirs, Conflict and Political Instability

Our theory is that the availability of male heirs affected the likelihood of conflict, andsubsequently development, in the (i) cultural context of political gender bias, and the (ii)institutional context of weakly specified succession rules. That is, if we view the availabilityof male heirs as being largely contingent (i.e., reflecting luck) then the the main theoretical

5

contribution of our work is that luck works in combination with both culture and institutionsin shaping development paths.

In nearly all of medieval Europe, inheritance practices contained a strong element of gen-der bias, preferring male heirs over female heirs, and male lines of descent over female lines.7

While matrilineal inheritance systems are common in some parts of Africa, Southeast Asiaand pre-Columbian America (Hartung, 1985), they were virtually unknown in pre-modernEurope, a continent that is also noted for having had a strong pro-male gender bias relative toother areas of the world (Boserup, 1970; Alesina, Giuliano and Nunn, 2013). European aris-tocracy felt strongly that women would be incapable of exercising military power, and that amarried woman would be heavily influenced, if not controlled, by her husband (McLaughlin,1990; Jansen, 2002). Moreover, certain polities of the former Carolingian Empire that prac-ticed “Salic law,” outright prohibited inheritance through female lines of descent (Herlihy,1962; Potter, 1937). Under these prejudices, it is not surprising that the availability of maleheirs would be important for guaranteeing peaceful and smooth successions.

In addition, and despite such cultural gender bias, political stability in medieval Europeanmonarchies was closely tied to the succession procedures of the regime, with clearer successionprocedures reducing instability (Herz, 1952). Kokkonen and Sundell (2014) have shown, forexample, that the transition to primogeniture-base inheritance systems in medieval Europewas associated with increases in leader tenure and overall political stability. They traceEurope’s transition from agnatic systems of succession and selection by assemblies to asystem of male-preferred primogeniture, which gives priority to the first born son of themonarch, but allows female(-line) heirs to inherit the throne when direct male heirs arenot available (Ward, 2014).8 They document, however, that no matter which system wasused, the nobility were rarely perfectly coordinated on the succession order, and the politywould experience a great deal of conflict and instability resulting primarily from competingclaims to succession. For example, the throne might be claimed by living daughters and theirhusbands, or more distant relatives, who would cite various conflicting inheritance rules asjustification.9 The Hundred Years War is perhaps the best-known historical example ofa succession related conflict. The central issue of the war was whether the French throne

7In fact, all European monarchies had some form of male-preferred inheritance up to 1980: no Europeanmonarchy practiced absolute (gender-neutral) primogeniture before 1980 (Corcos, 2012).

8Here, it is worth noting that even in the polities that used elections to select a new monarch, 47% ofsuccessors were close male relatives of the deceased monarch, as compared to 60% of successors in areas thatpracticed agnatic succession or male-preferred primogeniture.

9Even where all the competing claimants were female or female- line, the expectation that a queen’shusband would exercise substantial power, and the practice of marrying royal daughters into aristocratic androyal families, meant that there were frequently powerful in-laws with much stronger incentives to intervenethan they would have had if their were male heirs.

6

descended by primogeniture to Joan II of Navarre, half-sister of the deceased King John I (thePosthumous), and subsequently to Edward III of England; or whether Salic law prohibitedthis inheritance through a female-line, making John’s uncle, Phillip of Poitiers, the rightfulheir (Sumption, 2009).10 In fact, it is difficult to point to any part of the continent wherethe lack of a close male heir was not associated with a destructive war arising from suchcompeting claims. In the supplemental appendix, we provide a list of 18 known cases wherethe death of a king without any male heirs led to a civil war, in which a disputed successionwas unequivocally the main issue.

The importance of institutionalized succession rules points to the central place that co-ordination problems play in political organization: after a ruler dies, the nobles (and othermembers of the polity, more generally) must find a way to coordinate around a successorif they are to avoid conflicts arising from succession disputes (Kokkonen and Sundell, 2014;Tullock, 1987; Kurrild-Klitgaard, 2000). It also suggests that a shortage of male heirs shouldaffect conflict and instability less—and therefore have a smaller effect on contemporarydevelopment—in polities where succession norms are more strongly institutionalized. Wheresuccession rules are well-defined, the ambiguity in succession caused by a lack of males ismuch easier to resolve. For example, most regions in our study eventually institutionalizedmale-preferred primogeniture as their succession norm, and the effect of the availability ofmale heirs should be weaker in polities that were early adopters of this norm. In fact, inher-itance norms such as primogeniture might themselves have become stronger in places wherethe abundance of male heirs limited the frequency with which the norm was challenged, inturn reinforcing the norm. We explore a number of these implications in the later part of thepaper. However, the main implication of our theory, which we explore first, is that due tothe political gender bias against female(-line) succession, and the weak institutionalizationof inheritance practices that pervaded all of Europe in the Middle Ages, the availability ofmale heirs should have affected the the success of medieval state building efforts through itsaffect on conflict; and thus it should have had a long-run effect on development.

3 Contrasting Examples: Naples and France

One of the most striking examples of how violent conflict could destroy even a highly orga-nized polity comes from the experience of Naples. According to Takayama (1993), the Nor-

10The Hundred Years War thus illustrates how a disputed succession could result in destructive war wheninheritance norms are in conflict and it is unclear which norm trumps the other. In addition, because the warinvolved the kings of England prosecuting their claim to the Kingdom of France, it also demonstrates howthe tangled pattern of elite marriages could result in external rulers having plausible claims to the thronesof other monarchies.

7

man Kingdom of Sicily was one of the best governed medieval polities in the 12th century,with some already strong state institutions, such as a large bureaucracy and tax gatheringapparatus that drew on pre-existing Arab and Byzantine traditions. This view is consistentwith that of Abulafia (1988), who describes how the liber augustalis of Frederick II enshrineda set of limitations on aristocratic and urban power that were remarkable for their time, in-cluding a ban on wearing arms in public, a ban on the sale of fiefs, depriving the barons andtowns of the right to administer justice, and subjecting clerics to the royal courts.

In the next three hundred years, however, the Neapolitan state was comprehensivelydestroyed by a series of civil wars brought on by a shortage of male heirs. Six of its monarchsdied without sons during the Middle Ages, leading to three destructive periods of politicalinstability. These wars not only disrupted the operations of the central bureaucracy, butalso allowed the aristocracy to win back some of their autonomy (Jones, 2000). Femaleinheritance provides the backdrop to the final and most destructive series of conflicts, whichbegan in 1343 when King Robert died without any living sons, leaving the throne to his grand-daughter Joanna. This led to a series of conflicts between Joanna and the descendants of themore distant male line heirs, and efforts to resolve the uncertainty by marrying the queento one of them. These efforts were not successful: both of Joanna’s first two marriages weremarked by intrigues by the husbands against her, and contemporaries accused Joanna of themurder of her first husband, Andrew of Hungary. The result was a complicated series of atleast four revolts and pacifications intertwined with the rivalries of the Neapolitan nobility,as well as with the contending claims of two rival popes.11 The conflict lasted even afterJoanna was strangled in prison in 1381, by which time the division between supporters of herclaim—eventually vested in the kings of France, and the male-line Durazzo claim, eventuallyvested in the kings of Aragon—ran deep within the kingdom’s political class. In the nextcentury and a half, the two factions would conduct five successful coups, usually with thesupport of foreign money and mercenaries. During this period, the nobility regained mostof the autonomy that they had lost in the 12th and 13th centuries.12

The persistence of the institutional weakness in Southern Italy relative to Northern Italywell after 1500, and its close correlation with medieval political history, is a pattern for whichthere is a good deal of evidence, both quantitative and qualitative (Guiso, Sapienza and Zin-gales, 2016; Putnam, Leonardi and Nanetti, 1994). There are several plausible institutional

11These revolts took place in the periods 1347-1352, 1353-1360, 1379-1384, and 1387-1400.12Guiso, Sapienza and Zingales (2016) and Putnam, Leonardi and Nanetti (1994) trace the decline of

Naples to the rise of royal power relative to the urban republics of Northern Italy. We do not includenorthern Italy in our analysis, instead comparing to Naples to other monarchies. However, we believe thatthis hypothesis cannot explain outcomes outside Italy, since strong medieval monarchies are often associatedwith economic growth in other areas (Stasavage, 2014).

8

reasons for this: For instance, levels of social trust appear to be lower in the south (Putnam,Leonardi and Nanetti, 1994; Banfield, 1967), and lower trust is a well known legacy of vio-lent conflict (Colletta and Cullen, 2000) and extractive instability (Nunn and Wantchekon,2011). One other factor worth mentioning is the persistent political power of the aristocracywell into the 20th century, in particular their maintaining local political fiefs with little in-terference from central authority, much as they had in the medieval period. Astarita (2002,1), in his remarkable study of the “ ‘continuity of feudal power” in Brienza, found that thelocal lords for three centuries “owned land there, they levied fees, they administered justice,they controlled local trade, they were involved in every aspect of the life of their subjectpopulation,” and that feudal powers were more “widespread, pervasive and influential” inNaples and Sicily than in any other part of Italy (19).

The experience of Naples can be contrasted to the experience of the Ile-de-France, whichenjoyed the exceptional stability of its royal family for more than three centuries from 987-1316AD. Every Capetian king was succeeded by an adult son in this period, a run of geneticgood luck unparalleled in all of Europe. Historians, such as Lewis (1981), have argued thatdynastic stability was a key factor in the rise of the Capetians from a regional power in theIle-de-France to become the rulers of a centralized state covering much of Western Europe.Not only were the Capetians spared the problem of internal conflict, but the stability of theirdynasty put them in an advantaged position to expand their holdings by marrying their sonsto the female heirs of their neighbors. This enabled them to acquire the territories of severalrulers who had once rivaled them in wealth and power, including the Counts of Toulouse(1271) and Champagne (1314).

The territorial growth of the French royal domain was accompanied by a series of insti-tutional changes that took place within the domain as well. Baldwin (1991), for example,explains how 11th century France, like most medieval polities, contained a confusing massof conflicting political authorities—aristocratic, clerical and urban—all claiming the right totax, administer justice, make war and regulate trade. These claims were difficult for the royalauthorities to limit, both because of a lack of a permanent royal administrative presence inthe provinces, and because many of these petty lords possessed fortified bases. Baldwin doc-uments how in the 12th and 13th centuries, the Capetians, most notably Philip II Augustus,moved vigorously to limit the power of these rulers, seizing the castles of recalcitrant lords,basing permanent royal representatives (the ballis) in the major towns, and by institution-alizing a waiting period to prevent the escalation of private disputes into destructive warsbefore being diverted into the royal courts (via the quarantaine de la roi).

The end of French good luck came after Charles IV died leaving no close male heirs in1328. The resulting disputes for the French throne between the Dukes of Valois (Charles’s

9

closest male-line heirs), and kings of England (who held a closer, but female-line claim)resulted in the Hundred Years War. This was, nevertheless, the only failure in the male lineof the French monarchy between 987 and 1498AD.

The most persistent legacy of the power of the Capetian kings was a relatively centralizedstate based in Paris—a legacy that was not altered by France’s subsequent history. Since theirmedieval predecessors had already done much of the work of eliminating the power of feudalaristocrats and appenaged princes, it was easier for their Early Modern predecessors (notablyLouis XIV and Cardinal Richelieu) to complete their work, subordinating the aristocracyto the state by incentivizing them to be oriented towards the court and the royal servicerather than their estates and replacing their local authority by that of royal officials (Beik,1985). The relative ease of this achievement stems from the fact that the largest and mostindependent princes, such as the Valois Dukes of Burgundy, had been defeated centuriesbefore. Many other direct continuities can be traced between the relatively strong institutionsof the medieval French state and institutions of the contemporary one. Perhaps the mostinteresting is the French Army, which is descended directly from the creation of the modernwest’s first standing army, the compagnies d’ordonnance, by the Valois kings.

4 Data and Empirical Strategy

Ideally, we would test our theory by collecting a full dataset of succession disputes thatwere caused by the absence of a male heir, but it is impossible to assemble the universe ofsuccession disputes let alone code their main causes. Estimating the effect of these successiondisputes on contemporary development would then require us to find an exogenous source ofvariation in their likelihood since succession disputes are likely to be correlated with variousregion characteristics that directly affect development. In light of these limitations, wetherefore adopt a more indirect approach in which we estimate the reduced form relationshipbetween factors that drive succession disputes (mainly through the availability of male heirs)and contemporary development. We then present supporting evidence that the estimatedrelationship reflects the theoretical mechanism.

To measure contemporary development we use the Log of GDP per capita, adjustedfor purchasing power, and averaged across 2007-09. These data are from the Quality ofGovernment (QoG) EU Regional Database and are measured at the largest subnational(“NUTS”) level by the European Union (Charron, Dijkstra and Lapuente, 2014).13

13These data include only regions that are part of the EU, and so the regions in Turkey, Albania, Switzer-land, Norway and all of the former Yugoslavia apart from Slovenia are not included.

10

The majority of modern European regions can be associated with a single ruler for mostyears between 1000-1500AD. We collected data on 812 unique rulers that reigned in thisfive hundred year period, and associated each to regions for all years in which he or shecontrolled approximately all of the region that year.14 For each ruler, we collected data onthe numbers of legitimate and illegitimate male and female children, how many of each diedas infants, how many were alive at the time of the ruler’s death and on how each ruler wasreplaced. Our coding is based primarily on McNaughton (1973), Morby (1989), and theHull University Directory of Royal Genealogical Data. Royal genealogies are one of the fewdata sources of medieval society that were recorded will some regularity, and overall, we aremissing data on only 5.5% of region-years.

We focus on the period 1000-1500AD because data on the children of monarchs are scarceprior to 1000, and European polities already begin resemble much more institutionalizedterritorial states after 1500. These more institutionalized states, with their written successionlaws and control over the nobility, should be less vulnerable to instability caused by a lackof male heirs, though the shortage of male heirs remained a common (but, by comparison,less important) source of territorial conflict in Europe until the 19th century.15

We used these data to construct several measures of the likelihood of the availabilityof male heirs to the rulers of each region for the period 1000-1500AD. Perhaps the moststraightforward way is to consider the number of times a monarch died or was replacedwithout a (legitimate) male heir—an inverse measure of the independent variable. Anotherway is to take the total number of male heirs of a monarch that were alive at the timeof the monarch’s death or replacement, and sum this number over monarchs that ruled inthe five-hundred year period. A third way is to construct the weighted average number ofmale heirs to each monarch using the lengths of the monarchs’ reigns as weights. A fourthway is to consider the fraction of years that the region was ruled by a monarch who had aliving male heir. A fifth measure is the fraction of males among all legitimate children of allmonarchs that ruled the region in the five hundred year period. We refer to this measure as“percent male” and depict it in Figure 1. Finally, a sixth measure is the fraction of firstbornlegitimate children of a region’s monarchs that were male. All of these six measures arecorrelated (in the right way) with each other.

14The coding is based on the Atlas of Medieval Europe (Classen et al., 1997). We coded rulers as “ruling”a territory if they exercised autonomous political authority within the territory, even if they acknowledgedthe suzerainty of another ruler, such as the French king or in the Holy Roman Emperor. These larger rulerswere coded as only ruling their own demesne territories. Readers interested in the linking of regions anddynasties should consult the comprehensive listing provided the supplemental appendix.

15Examples include the wars of the Spanish Succession (1701-14) and Austrian Succession (1740-48).

11

Figure 1. Shaded regions show the proportion of males among the legitimate children of rulersin the period 1000-1500AD. White regions either fall below our 200 year cutoff, or are regionsin non-EU states (Turkey, Norway, Switzerland).

Each of these measures has advantages and disadvantages. While the number of monarchswho died without heirs closely captures our theory, it is endogenous to royal birthrates anddeath rates, and thus to the existing characteristics of regions. Percent male, which is anoisier measure of the treatment variable, is only influenced by the gender ratio at birth.It should be noted, however, that all of these measures produce similar empirical results.For similar reasons, we have not discussed two additional variables that are endogenous tolocal political patterns, the availability of brothers, and the age of monarchs when theyinherited.16

Our empirical approach assumes that political actors across Europe faced fairly similarpolitical conflicts. Therefore, to create our baseline sample, we excluded the following region-years from the data: (i) region-years in which the region was controlled by multiple rulers,usually because they were divided among several petty lords, or because the modern politicalboundary cuts across a medieval one, (ii) region-years in which the region was not ruled bytraditional monarchs, but rather by the church or by urban republics, or was populatedby unorganized tribal groups, and (iii) region-years in which the region was controlled byMuslims, who had very different inheritance and marriage practices than non-Muslims inEurope (Blaydes and Chaney, 2013).17 After excluding these region-years, some regions

16In addition to endogeneity, the effect of the availability of brothers is theoretically unclear, since thereare many cases of them claiming the throne in opposition to nephews, or each other.

17The exclusion of the urban republics, which are primarily located in central and northern Italy, isimportant to note since several influential accounts of the long term influence of medieval politics have

12

have very few years of data, so we excluded regions with less than 200 years of data toavoid basing our inferences on regions where the consequences of (Christian) monarchy werehistorically unimportant. In the appendix, we show that our results are robust to varyingthis threshold.

Finally, in arguing that the long-term consequences of medieval inheritance practiceshave persisted over time, we have implicitly assumed that there is some level of continuityin the population of European regions between that time and ours. In seven regions ofcontemporary Poland and the Czech Republic, we know this assumption to be false.18 Theseregions were populated by German speakers (and ruled by Germanic rulers during the MiddleAges) but their entire populations were forcibly removed after World War II and replaced byPoles and Ukrainians (Schechtman, 1953). Given a population change of nearly 100%, wehave no reason to expect that medieval events could affect contemporary outcomes in theseareas by either cultural or institutional means. Therefore, we exclude these regions from themain analysis. This gives us 114 regions, which constitute our baseline sample. These, alongwith the seven omitted regions, are depicted in Figure 1. Summary statistics for all of thedata used in this paper are provided in Appendix Section A.1.

5 Results

5.1 Baseline Estimates

Table 1 presents coefficient estimates from OLS regressions of the effect of each of our sixmeasures of the independent variable on contemporary GDP per capita. In addition tothe conventional standard errors, Table 1 also reports standard errors clustered at the levelof what we call a “macro-polity.” Since our unit of analysis is defined by modern regionboundaries, these modern regions cannot be considered the unit of treatment. In particular,single medieval rulers often ruled over multiple modern-day regions, suggesting that thetreatment was assigned at units larger than the modern region. For example, the threeobservations in Walachia that enter our dataset all have the same values of our independentvariables because all three of these regions were ruled in the medieval period by the samesequence of monarchs. The same is true for the eight observations in Hungary and the fiveobservations in Denmark. Even areas with different sets of monarchs in different periods

focused on the positive impact of these independent city states (Guiso, Sapienza and Zingales, 2016; Putnam,Leonardi and Nanetti, 1994). Since we are excluding the regions of Europe with the highest levels of wealthand political participation, our estimates of the long-term persistence of medieval political patterns areprobably more conservative than they otherwise would be.

18These are Dolnoslaskie, Jihozapad, Lubuskie, Opolskie, Pomorskie, Severovychod and Severozapad.

13

Table 1 – Baseline Results with Alternative Independent Variables

Dependent variable is Log of GDP per capita averaged 2007-09

Total Weighted ave. # of Zero % Yr with FirstbornMale Heirs # Male Heirs Male Heirs Male Heir % Male % Male

(1) (2) (3) (4) (5) (6)

Coef. Estimate 0.012 0.248 −0.041 1.647 7.308 1.317(0.004)∗∗ (0.079)∗∗ (0.012)∗ (0.502)∗∗ (1.118)∗∗ (0.317)∗∗

[0.006]† [0.131]† [0.022]† [0.826]† [2.894]∗ [0.584]∗

{0.005}∗ {0.146}† {0.021}† {0.764}∗ {3.028}∗ {0.570}∗

Estimate of 1 s.d. ↑ 20.5% 20.0% −17.9% 21.4% 41.0% 26.8%Largest 95% C.I. [0.4%, 44.8%] [−2.7%, 48.3%] [−33.2%, 0.9%] [0.3%, 46.9%] [6.7%, 86.4%] [3.2%, 55.8%]

N 114 114 114 114 114 114R2 0.085 0.080 0.091 0.072 0.276 0.133†p < .10; ∗p < .05; ∗∗p < .01

Note: OLS estimates of the effects of various independent variables on Log of GDP per capita adjusted forPPP and averaged between 2007 and 2009. The independent variable in column (1) is total male heirs, incolumn (2) is the weighted average number of male heirs to each monarch using the lengths of the monarchs’reigns as weights, in column (3) is the number of times a monarch died or was replaced without any maleheirs, in column (4) is the fraction of years ruled by a monarch who had at least one male heir at the time ofdeath or replacement, in column (5) is percent male, and in column (6) is the percent of males among onlyfirstborn children of monarchs. Standard errors in parentheses. Standard errors clustered at the macro-politylevel in square brackets. Block bootstrapped standard errors in curly brackets.

are not fully independent of each other because of their shared membership in supra-localpolitical units like the Holy Roman Empire, “France,” or “Poland.” Therefore, to partiallyaccount for this dependence, we classified every modern region as belonging to a particular“macro-polity” that existed at the start of the period of our analysis.19 The key upside toclustering errors at the macro-polity level is that the borders of these macro-polities arestable, unlike the borders of the medieval polities (see Section 5.2); and, although differentrulers ruled over different sets of regions, most of them ruled over regions that belong to asingle macro-polity. The downside is that even the macro-polities are not exactly the unit oftreatment, and we have only 14 clusters so even if the macro-polities corresponded perfectlyto the unit of treatment, these standard errors may be downward biased. Therefore, topartially address this concern Table 1 reports standard errors estimated using the blockbootstrap method as well (Angrist and Pischke, 2008, p. 319-23).

19These macro-polities are Aragon, the Byzantine Empire, Castile, Denmark, England, France, the HolyRoman Empire, Hungary, Poland, Portugal, Scotland, Sicily, Sweden and Walachia. While the formalboundaries of “France” did not correspond to the territory ruled by French kings until after 1500, areaswithin these boundaries came to share a deference to the Parisian court well before 1500, which shaped theirpolitical experiences. Some of these polities were divided into warring minor units, but their boundariescontinued to provide the structure for political contention.

14

In what follows we adopt column (5) of Table 1 as our “baseline specification.” Thechallenge with the independent variables used in the first four columns is that they may varysignificantly with unobservable characteristics of regions that affect family size and survivalrates, which in turn may directly affect development. The measures used in columns (5)and (6) are much less likely to be affected by these confounders. For example, we show inSupplemental Appendix A.1 that percent male variable is balanced across a variety of regioncharacteristics.20 We also provide some additional evidence in Supplemental Appendix A.2that the effect of environmental factors on sex ratios is unlikely to be accounting for theeffect of percent male on contemporary development.21 Nevertheless, between the last twomeasures (percent male and firstborn percent male) we face a tradeoff. Whereas firstbornpercent male cannot be affected by endogenous stopping rules, percent male can—though itis unlikely that it would be heavily affected since it is an average over many children. Onthe other hand, percent male is a less noisy measure of the availability of male heirs thanis firstborn percent male, which accounts for the fact that the measured effect in column(6) is attenuated in comparison to that of column (5).22 Since we find limited evidence forstopping rules in Supplemental Appendix A.3, we resolve this tradeoff in favor of the percentmale measure. Figure 2 shows the bivariate relationship between percent male and GDP percapita with units labeled by their NUTS region code.

Table 1 indicates that no matter how we measure it, the likelihood of having male heirsappears to have a substantial—albeit, very imprecise—positive effect on contemporary GDPper capita. For example, the point estimate for the effect of a one standard deviationincrease in percent male (roughly 0.047) is a rise in GDP per capita of 41.0%, but it is moreimportant to emphasize the large the 95% confidence interval for this estimate (taking thelargest estimate of the standard error) of [6.7%, 86.4%]. Similarly, a one standard deviationincrease in firstborn percent male (roughly .180) is associated with a rise in GDP per capitaof 26.8%, again with a large 95% confidence interval of [3.2%, 55.8%].

20These include urban density in the year 1000, history of Muslim rule, presence of pagan tribes, use ofelected monarchy, whether the region belonged to the Roman empire, whether it belonged to the Carolingianempire, whether it has a coast, or is on the Atlantic coast, the age of the state, and heating degree days.

21It is well known in the demography literature that sex ratios can be affected by environmental factors.For example, the economic stress caused by German re-unification caused the proportion of male births torise by 0.004. However, the difference in our percent male variable between the richest and poorest quartilesof European regions is much larger at 0.047. Figure A.1 in the supplemental appendix provides additionalsuch evidence that our results are probably not being driven by the effect of environmental factors andparental traits on sex ratios.

22Note that, while the number of monarchs’ children in the dataset vary somewhat from region to region,differences in these sample sizes only result in measurement error in the independent variable. Note alsothat the use of the 200 year threshold means that only four regions had fewer 40 child observations, 80%had more than 60, and the region with the most had 145.

15

AT11

AT12

AT13

AT21AT22AT31AT33

BE1

BG31BG32

BG33BG34

BG41

BG42

CZ01

CZ02CZ03CZ04CZ05

CZ06CZ07

DE1 DE2

DE3DE4

DE7

DEDDEG

DK01

DK02

DK03DK04DK05

ES11ES12

ES13

ES21ES22ES23 ES24

ES30

ES41

ES42ES43

ES51

ES52ES53

ES62

FR10

FR21FR22

FR23FR24FR25FR26 FR41FR43 FR52 FR53

FR61 FR62FR63FR81

FR82

GR1GR2

GR3GR4

HU1

HU2

HU3

ITC1ITC2ITD1

ITF1 ITF2ITF3ITF4

ITF5ITF6ITG1

NL2

NL3NL4

PL11

PL12

PL21

PL31PL32PL33

PL41PL43

PL51

PL52PL61PL63

PT11

PT15

PT16

PT17

PT18

RO11RO12RO42

SE1

SE3

SK01

SK02SK03SK04

UKCUKDUKEUKFUKG

UKH

UKI

UKJUKK

UKL

UKM

89

1011

Log

GDP

per

cap

ita, P

PP, a

vera

ged

2007

-200

9

.4 .45 .5 .55 .6Fraction of Males among Monarchs' Children, 1000-1500AD

Figure 2. The fraction of males among all legitimate children of monarchs in the period 1000-1500AD plotted against Log of GDP per capita adjusted for purchasing power and averagedbetween 2007-09. The unit of data are NUTS regions, and data points are labeled by theirNUTS code.

The relatively large size and lack of precision in our estimates (due to low statisticalpower) is an inherent limitation of this empirical exercise. If we add the pre-treatmentcovariates mentioned in footnote 20 to the specification in column (5), the estimate of theeffect of a one standard deviation rise in percent male drops to 29.1% with a 95% confidenceinterval of [5.0%, 58.8%].23 Alternatively, if we add modern country fixed effects to thisspecification, the effect of a one standard deviation rise in percent male drops is a moremodest 7.8% with a 95% confidence interval of [0%, 16.6%].24 We should note, however, thatthese estimates potentially suffer from serious post-treatment bias since the map of Europechanges frequently and very dramatically after 1500, and these changes may be correlatedwith region characteristics.

5.2 Alternative Samples and Specifications

In Supplemental Appendix A.4, we report results from an alternative analysis where weredefine the unit of analysis to be the “medieval polity” in particular year. A medievalpolity in year 1000, for example, consists of the area that was ruled by a particular ruler in

23The coefficient estimate is 5.437 (s.e. = 1.080, macro-polity clustered s.e. = 2.245, block bootstrappeds.e = 2.124, N = 107, R2 = 0.520).

24The coefficient estimate is 1.594 (s.e. = 0.850, N = 114, R2 = 0.879).

16

year 1000. Medieval borders are changing rapidly and dramatically over the five-hundred-year period of our analysis. To define the borders of a medieval polity, we need to fix aparticular year, which dramatically reduces the number of observations for any particularyear since most rulers controled multiple regions. Despite this, our percent male variable hasa discernible effect on contemporary GDP per capita when we redefine the unit of analysis tobe the medieval polity by fixing political boundaries at the turn of each century from 1000-1500AD. In addition to this, we also report the very similar results of a specification wherewe use “treatment units” (the largest clusters of geographically contiguous NUTS regions forwhich the percent male variable does not vary) as the units of analysis.

In Supplemental Appendix A.5, we report some additional robustness checks. First, asFigure 1 indicates, there are substantial differences in our main independent variable acrosslarger regional blocs, raising concerns that our estimates are being driven by particularclusters of regions. For example, most of Poland is poorer than most of France, and hasmuch lower values of percent male. We address this concern by redoing the analysis afterremoving various various geographic clusters of macro-polities, and showing that the effectof percent male on contemporary development is positive and relatively stable across thesesubsamples. Second, we show that our results are robust to including the seven omittedregions of Poland and the Czech Republic. Third, we show that they are robust to varyingthe 200 year cutoff that defines the baseline sample. Fourth, we show that our results arerobust to removing capital regions from the analysis. Fifth, we show that the results arerobust to removing foreign rulers, suggesting that our results are not driven by the possibilitythat monarchs with more male heirs conquered more polities, especially richer ones, and arethus over-represented in the data. Finally, we show that our results are robust to removingthe seven years of the Black Death (1347-1353) from the region-years.25

5.3 Multiple Male Heirs

In Supplemental Appendix A.6, we investigate whether having multiple male heirs is harmfulfor development, because perhaps it leads these heirs to conflict over succession. We findno support for this hypothesis: having multiple as opposed to only one male heir is not anymore harmful for development, and may be beneficial. The result is not surprising, given thatmost polities had relatively clear succession orders for close males relatives of the monarchregardless of the succession rule. For example, in polities that used primogeniture, themonarch’s oldest male son is first in the line of succession, followed by the second oldest maleson, etc. Each potential male successor is likely to have known his position in the succession

25The Black Death, a calamitous event that result in the loss of a third of Europe’s population could haveaffected regions differently, particularly hurting already poorer regions.

17

hierarchy. Nobles would have clear and coordinated expectations regarding the successionhierarchy as well. Where a female or female-line heir fell in the order of succession was, onthe other hand, much less clear. Therefore, in the absence of close male relatives, successiondisputes could be common, but they were unlikely to be common when the monarch hadmany potential male successors.

5.4 Illegitimate Children Placebo Test

Since norms against succession by illegitimate children were very strong—though brokenin a few notable cases by dynamic men—we have no reason to think that the sex ratio ofillegitimate children should have influenced medieval politics.26 Though illegitimate childrenare likely to be underreported, we have data on several hundred in our period. We ran an OLSmodel of the effect of the percent of illegitimate male children among all illegitimate childrenof European monarchs that ruled a particular region of Europe between 1000-1500AD on ourmain dependent variable, Log of contemporary GDP per capita—the same specification asin Table 1 column (5), but with percent male among illegitimate children as the independentvariable. The coefficient estimate in a sample only slightly smaller than our baseline samplewas small, barely discernible, and actually negative at −0.504 (s.e. = 0.273, macro-polityclustered s.e. = 0.506, block bootstrapped s.e. = .522, N = 107, R2 = 0.031). This providesadditional justification to our claim that the large positive results we find in our baselineestimates are probably not being driven by underlying environmental differences.

Although royal genealogies are some of the best sources of data from the medieval period,one may be concerned that biased recording in the numbers of legitimate males may bedriving our results. For example, rulers in already stronger states may have been morecapable of “legitimizing” an illegitimate heir, or claiming a son that is not theirs. The datado not show signs of biased recording, however. For example, the monarchs in our data hadmale children 50.9% of the time, corresponding to a ratio equal to that of the contemporaryWestern world (Grech, Savona-Ventura and Vassallo-Agius, 2002).

Nevertheless, to address the concern that biased recording is driving our results, we reportthe results of a sensitivity analysis in which we continued to randomly delete legitimate maleheirs of rulers that ruled regions with higher than median contemporary GDP per capitauntil the effect of percent male on GDP per capita in the specification of column (5) in Table1 lost significance at 10%. We did this a hundred times. The mean number of deletionsneeded was 397 (s.d. = 10.25, min = 360, max = 430), or 14.5% of male heirs. Thus,we estimate that more than one in seven legitimate princes in the wealthiest regions would

26William the Conqueror, for instance, was an illegitimate son of Robert I, Duke of Normandy. On theother hand, there is no example of succession by or through an illegitimate daughter in our data.

18

have to be illegitimate for our results to be explained by biased recording. This is a fairlyconservative analysis since it assumes that only the monarchs of the wealthiest regions couldlegitimize their illegitimate heirs— i.e., none of the monarchs of the poorest regions could.

5.5 Luck

Our interpretation of the effect of percent male on contemporary development is that “luck”has played an important role in setting development paths. The balance tests and evidenceon environmental factors reported in Supplemental Appendix A.1 and A.2 support thisinterpretation. We now report some additional evidence that the percent male variable canbe interpreted as luck.

Fixing the total number of children of each ruler, we randomly and independently drewa new sex for each child of each ruler, with the probability of a male draw fixed at 0.51.This number corresponds to the modern sex ratio at birth estimated for North America andEurope, where sex-selective abortions are much rarer (almost nonexistent) in comparison tothe rest of the world (Grech, Savona-Ventura and Vassallo-Agius, 2002). We then computedvalues of percent male for each region from the simulated data. For each of 1000 such simu-lations, we then did a Kolmogorv-Smirnov (KS) test of the hypothesis that our empiricallyobserved distribution of sex ratios in European regions are from the same distribution asthe simulated sex ratios. Only 6.4% of p-values were below 0.10 and only 2.4% were below0.05.27 This suggests that it is difficult to reject the hypothesis that the distribution ofpercent male in the data is consistent with luck. In fact, Section A.1 of the supplementalappendix presents a figure depicting the cumulative distributions for the 1000 simulationsalong with the actual data. The figure shows that the observed distribution looks very muchlike a typical draw of the randomly generated distributions, suggesting that the distributionof percent male is consistent with random draws, i.e. luck.

The idea behind these tests is that the distribution of percent male in the data wereconsistent with luck then it should closely resemble the randomly generated distribution ofpercent male in which, fixing the total number of polities and total number of children ineach polity, male children are drawn with the same probability that they are likely to beborn in the absence of sex-selective abortions. The KS test asks how close these distributionsare, and we ask this question over 1000 different draws of the random distribution. If theobserved distribution is consistent with luck, it should look like a typical draw from themany that we drew; and very few of the p-values for the 1000 KS tests should be small.

279.7% were below 0.15, 15.2% were below 0.2 and 63.4% were below 0.5.

19

Table 2 – Tracing the Effects Through Time

Dependent variables are Log of Urban Density, various years

1000 1300 1400 1500 1600 1700 1800

(1) (2) (3) (4) (5) (6) (7)

% Male 1.541 3.371 3.505 3.402 3.529 4.286 4.982(1.103) (1.567)∗ (1.705)∗ (1.765)† (2.105)† (2.327)† (2.563)†

[1.195] [1.804]† [1.370]∗ [2.143] [2.006] [2.371]† [3.097]{1.309} {1.989}† {1.479}∗ {2.083} {2.316} {2.142}∗ {2.975}†

N 114 114 114 114 114 114 114R2 0.017 0.040 0.036 0.032 0.025 0.029 0.033†p < .10; ∗p < .05; ∗∗p < .01

Note: OLS estimates of the effects percent male on Log of Urban Density in the years 1000, 1300, 1400,1500, 1600, 1700 and 1800 in columns (1)-(7) respectively. Urban Density is defined as (1+urban population)divided by area of the region in square km. Standard errors in parentheses. Standard errors clustered at themacro-polity level in square brackets. Block bootstrapped standard errors in curly brackets.

6 Evidence for the Mechanism

6.1 Persistence of the Effects Through Time

If the availability of male heirs in the medieval period has had a persistent effect on develop-ment then we should be able to trace the effects of our main explanatory variable, percentmale, on measures of economic development over time. Direct measures of developmentare generally not available prior to the late 20th century, so we use urbanization data fromBairoch (1991) to measure development, which is the standard practice in the literature(see, e.g., DeLong and Shleifer, 1993; Acemoglu, Johnson and Robinson, 2002). Table 2shows that there is essentially no relationship between percent male and urban density inthe year 1000, as our theory would predict. However, in the year 1300 we begin to see apositive relationship, which gets stronger in subsequent centuries, implying a growing diver-gence in economic prosperity over time. These increasing effects are consistent with theoriesin which shocks affect trajectories rather than levels of development. The estimated effectgrows steadily from the 14th century to the 19th century, where it is about three fifths of theeffect size that we computed in our baseline specification in column (5) of Table 1. However,again we caution by emphasizing the precisions of these estimates, noting that the effectsget noisier over time. This is consistent with the idea that other intervening shocks havemade the effects of percent male less precise over time.

The persistent effects of our percent male measure should not be interpreted to mean thatthe numerous other events that shaped Europe between 1500 and 2009 were unimportant—infact, the imprecision of our estimates strongly indicates that they were not. The fairly wide

20

dispersion of GDP outcomes between countries within similar medieval inheritance values inour data indicate that medieval history is hardly the only important factor—in particular,communist rule during the cold war and access to Atlantic trade appear strongly correlatedwith contemporary GDP, at times mitigating or reversing medieval legacies. Rather, theevidence supports the more modest statement that these events have not been sufficient tocompletely obliterate the influence of events in the long past. In fact, relative to the verystrong claims for historical persistence being made in some branches of the political economyliterature (e.g. Alesina, Giuliano and Nunn, 2013), the claim that a correlation of this naturecan persist for five centuries seems quite modest.

What might be the specific institutions that allow this persistence, even long after thephysical damage from succession conflicts is rebuilt or the monarchies involved overthrown?There are several plausible pathways, and we are wary (pas Austin, 2008) of “compressing”historical periods and cases with distinct causal mechanisms. However, it is worth mentioningone that we believe is especially plausible: the presence of small, autonomous jurisdictionsbelow the level of the state. Examples might include seignories with autonomous taxationand judicial powers, lordships and city states with the power to levy customs duties, priv-ileged merchant groups with monopoly rights, or royal appenages not paying taxes to thecenter. Strong medieval monarchs were able to limit the legal powers of these “subnational”political actors, or even eliminate them entirely. Weaker monarchs, by contrast, were toofocused on political survival to destroy such powerful vested interests, and indeed might evenbe indebted to them for support during civil conflict. These subnational actors, each focusedon narrow rentseeking, are widely thought to be a major cause of the low levels of economicdevelopment of specific European regions (North, 1973) that could only be partially amelio-rated through very radical reforms (Acemoglu et al., 2011). While Early Modern monarchsmade efforts, sometimes successful, to eliminate these political units (Anderson, 1979), theywere hampered by path dependence—their medieval predecessors, had created institutionalguarantees such as charters and representative institutions that made eliminating these in-stitutions much more difficult than it would have been had they never been created in thefirst place. This is one reason why many medieval political patterns were able to persist forcenturies until disturbed by major external shocks (Acemoglu et al., 2011).

6.2 Coups, Conflict and Quality of Government

Our theory says that the likelihood of having male heirs affects the occurrence of conflictand political instability, which in turn affect contemporary development via state-building.Table 3 provides evidence for the first link in this mechanism. Column (1) of the upper panel

21

Table 3 – Male Heirs, Coups and Conflicts

Coups per year Wars per year

(1) (2)

% Male -0.053 -0.584(0.014)∗∗ (0.224)∗[0.038] [0.279]†{0.036} {0.281}∗

N 117 86R2 0.105 0.075

Ascension by coup War

(3) (4)

Lagged Legitimate Male Heirs -0.647 -0.106(0.126)∗∗ (0.024)∗∗

Region Fixed Effects No YesDecade Fixed Effects No YesObservations 1,399 51,349

† p<0.1, ∗ p<0.05, ∗∗ p<0.01

Note: OLS estimates of the effects of percent male on the number of violent transitions per year and thenumber of internal wars per year in columns (1) and (2) respectively. Column (3) and (4) in the lowerpanel are estimates from a panel logistic regression of the effect of the lagged number of legitimate maleheirs on whether or not ascension by the next monarch involved a violent transition, and the effect on warsrespectively. The unit of analysis in model (3) is the monarch-polity, and in model (4) is the region-year.Standard errors in parentheses. Standard errors clustered at the macro-polity level in square brackets. Blockbootstrapped standard errors in curly brackets.

shows that percent male has a negative effect on the frequency with which monarchs thatwere removed from power in a coup in a given region—either by being killed or by beingforcibly deposed—which we treat as a proxy for political instability. Column (2) shows thatpercent male has a negative effect on the frequency of internal wars.28 The internal war datais based on all wars mentioned in Kohn (2013).29 To distinguish internal wars (hypothesizedto decrease state capacity) from external ones (hypothesized to increase state capacity) wecode internal wars as those where fighting occurred within the territory, no matter whothe participants were. For instance, England is coded as not being at internal war during

28In a related paper Dube and Harish (2015) use sex of the first-born child as an instrument for whether ornot a European polity was subsequently ruled by a female monarch to estimate the effect of female rulers onthe likelihood of external conflict, whereas here we find that the presence of male heirs affects whether or notthere was an internal war in the polity. Our finding, however, does not speak directly to their identificationstrategy since their sample begins after our sample ends, and because it is possible that the effect of percentmale on external conflict works exclusively through the sex of the monarch.

29We combined information from Kohn (2013) with information from Classen et al. (1997) and modernregion boundaries to code these wars.

22

Table 4 – Coups, Conflicts and Development

Log GDP per capita Quality of Government

(1) (2) (3) (4)

Coups -43.795 -74.930(6.850)∗∗ (11.593)∗∗[10.577]∗∗ [26.421]∗{16.314}∗∗ {34.018}∗

Wars -1.222 -1.847(0.323)∗∗ (1.049)†[0.387]∗ [1.928]{0.405}∗∗ {1.740}

N 114 83 114 83R2 0.267 0.150 0.265 0.037† p<0.1, ∗ p<0.05, ∗∗ p<0.01

Note: OLS estimates of the effects of the number of coups and the number of wars on measures of contem-porary development. The development measure in columns (1) and (2) is Log of GDP per capita, adjustedfor PPE and averaged 2007-2009. In columns (3) and (4) it is a composite measure from the QoG EURegional Database containing three sub-composites: “Quality,” “Impart” and “Corrupt,” all of which arecontemporary measures based on surveys conducted in the last ten years. “Quality” is a composite measureof the quality of healthcare provision, public education and law enforcement, the extent to which corruptionexists in regional elections, and the extent to which the media reports corruption by politicians. “Impart” isa measure of impartiality in the provision of heath, education and law enforcement. “Corrupt” is a measureof the control of corruption in the provision of health, education and law enforcement as well as perceivedcontrol on corruption in the public sector. Standard errors in parentheses. Standard errors clustered at themacro-polity level in square brackets. Block bootstrapped standard errors in curly brackets.

the Hundred years war, since the fighting was confined to France.30 Since Kohn’s data forEastern Europe are very poor, we drop Eastern Europe when we use war as a dependent orindependent variable. Column (3) shows that in a panel logistic regression with monarch-polity as the unit of analysis, the estimated effect of the lagged number of legitimate maleheirs on the occurrence of a violent transition is negative. Column (4) shows that the sameis true with internal war as the dependent variable, and region-year as the unit of analysis.31

Table 4 provides evidence for the second and third links in the mechanism. Columns (1)and (2) show that the number of medieval coups and internal medieval wars are negatively

30We use this coding because conventional civil war coding rules require distinctions between state andnon-state actors that are anachronistic in this period. That said, our results are similar even if we code aninternal war as one in which there is no foreign involvement; for example, see the next footnote.

31If we replicate this model using a definition of internal war as one in which there is no foreign involvement(defined as intervention by the ruler of another territory), the results are similar: the coefficient on laggedlegitimate male heirs is −0.397 (s.e. = 0.0933, N = 29, 546). Similarly, if we use the predecessor’s numberof male heirs as the independent variable instead, then the coefficient estimate is −0.401 (s.e. = 0.034,N = 51, 338) while if we use the number of male heirs left by the predecessor that were above 18 years of agethe coefficient estimate is −0.173 (s.e. = 0.033, N = 51, 338). These alternative specifications are motivatedby the idea that if a monarch dies without leaving a (mature) male heir, this may adversely affect the reignof the successor throughout the entire period of reign—an idea that is not fully captured in column (4),which uses the one-year lagged number of legitimate male heirs to the throne as the independent variable.

23

Table 5 – State Development for Three Polities in 1500

Polity 1500 Revenues % Male Coups Wars GDP pc QoG

Castile 2.56 .491 2.27 0.207 $23,857 -0.610England 3.47 .519 7.70 0.190 $30,135 1.184France 5.68 .551 2.54 0.089 $27,167 0.354

Note: Various statistics for the three polities for which we have state revenue data from 1500 and exchangerate estimates. The 1500 revenue data were collected by Bonney (1995) and are reported in the EuropeanRegional Finance Dataset. We converted the data from local currency into ducats (Venetian gold coins)and normalized by the area of the polity in 1500. Percent male is an average across the regions of thepolity. Coups is the average number of coups that occur across the regions of the polity. Wars is the averageproportion of region-years in which an internal war was fought. GDP per capita is modern income estimatedin 2009 for the modern state or region that corresponds to the 1500 polity. QoG is the average Quality ofGovernment measure described in the note below Table 4.

correlated with contemporary GDP per capita. Columns (3) and (4) show that they arenegatively correlated with contemporary Quality of Government, which is a measure of thequality of contemporary state institutions from the QoG EU Regional Database (see thenote under Table 4 for details). This provides some plausibility to the second and third linksin our mechanism. Because we do not have measures of state quality from the medievalperiod with wide enough coverage, we are not able to provide any more direct evidence forthe second link. However, Table 5 presents data on state revenues collected in 1500 for threepolities for which we have revenue data and exchange rates (so as to make them comparable).Stronger states were able to collect larger revenues. The data suggest that the Castilian statewas weaker than the English or French states, and had more frequent internal wars, thoughit had fewer coups than did England. Our measures of contemporary development (GDPper capita and Quality of Government) are also lower in Castile than they are in England orFrance. Thus, the data are largely consistent with the idea that internal political conflictsare associated with lower levels of medieval state development, and greater medieval statedevelopment is positively associated with contemporary development.

6.3 The Institutionalization of Inheritance Norms

Different areas of medieval Europe had different inheritance norms, and different levels ofinstitutionalization of these norms. For example, many areas of the former CarolingianEmpire practiced Salic law, which prohibited inheritance through female lines of decent.Polities that were not part of the former Carolingian Empire may have had equally highamount of un-coordinated, or un-institutionalized, cultural gender bias against female-lineinheritance, but they did not institutionalize a tradition that outright prohibited it. Here,

24

we are interested in understanding how the effects of our percent male variable vary withthe degree of institutionalization of inheritance norms.

For the case of Salic law, theory is silent as to whether we should expect percent maleto have a higher or lower effect on contemporary development in Salic law regions thanelsewhere. If cultural gender bias was much higher in Salic law regions than elsewhere, violentconflicts resulting from succession disputes could be more common in Salic law regions whenpercent male is lower and close male relatives are scarcer. Alternatively, because successiondisputes often arose from female-line heirs prosecuting their claims to the throne, violentconflict might actually be rarer in regions that practiced Salic law because female-line heirsin these regions would less frequently dispute successions by more distant male-line relativeswhen close male relatives were not available. In this case, percent male might have a smallereffect on development in areas that practiced Salic law, because gender bias is so highlyinstitutionalized that conflict between male- and female-line heirs is rarer. The results ofcolumn (1) in Table 6 support the second perspective: percent male has essentially no effecton contemporary development in areas that practiced Salic law, whereas it has a very largeeffect on contemporary development in other regions.

Another instance of variation in the degree of institutionalization of succession rulescomes from comparing the tribal areas that practiced Germanic law to those that wereexposed to Roman legal traditions. Because Germanic law tended to be less codified thanRoman law, we would expect to see the effect of percent male be lower in the non-tribal areasthat were exposed to Roman law than in the tribal areas that practiced Germanic law.32

This hypothesis is confirmed by column (2) of Table 6, which shows that in the non-tribalpolities of medieval Europe that had Roman and Greek traditions, the effect of percent maleis significantly lower than it is in the tribal areas that had more weakly institutionalizedsuccession rules. Again, the historical evidence points to political instability being greater inthe tribal areas than in polities that drew on their Roman and Greek traditions to establishcoordinated rules governing succession when close male relatives were unavailable.33

A third instance of variation in the degree of institutionalization of inheritance normscomes from comparing monarchies that used primogeniture as their succession norm formore years to those that use primogeniture in fewer years. According to Kokkonen andSundell (2014) and Blaydes and Chaney (2013), European polities gradually began to adopt

32The tribal regions, like the regions that practiced Salic law, had more precise and more deeply specifiedsuccession rules that their citizens and nobles could rely on even when the monarch had no close malerelatives. However, this is very much a distinct analysis than the analysis of Salic law versus non-Salic lawregions, as evidenced by the fact that the 42 Salic law regions in our sample are exactly evenly dividedbetween the sets of 54 non-tribal polities and 63 tribal polities.

33For example, the average number coups from 1000-1500AD for tribal areas is higher, at 4.746, than fornon-tribal areas, at 3.056, with the difference being statistically significant at the 1% level.

25

Table 6 – Interactions with Inheritance Norms and Women Rulers

Dependent variable is Log of GDP per capita

Z = Salic Nontribal Primog. Women

(1) (2) (3) (4)

% Male 8.891 12.312 11.829 7.436(1.261)∗∗ (1.495)∗∗ (1.551)∗∗ (1.427)∗∗

[3.632]∗ [4.617]∗ [3.747]∗∗ [3.084]∗

{3.860}∗ {5.825}∗ {4.770}∗ {3.448}∗

Z 4.883 5.110 5.973 0.055(1.203)∗∗ (1.083)∗∗ (1.387)∗∗ (0.564)[1.865]∗ [2.743]† [2.567]∗ [1.009]{2.008}∗ {3.899} {3.573}∗ {1.458}

% Male × Z −8.750 −10.033 −11.634 −0.067(2.314)∗∗ (2.117)∗∗ (2.701)∗∗ (1.121)[3.648]∗ [5.219]† [4.754]∗ [1.990]{3.991}∗ {7.479} {6.611}† {2.974}

N 114 114 114 114R2 0.414 0.399 0.381 0.277

†p < .10; ∗p < .05; ∗∗p < .01

Note: OLS estimates of models interacting percent male with other variables, with Log of GDP per capitaas the dependent variable. Column (1) is the interaction with regions that used Salic law, column (2) withnon-tribal regions that adopted Roman law as opposed to Germanic law, column (3) with the fraction ofyears that the region used primogeniture as its succession rule, and column (4) with number of women rulers.Standard errors in parentheses. Standard errors clustered at the macro-polity level in square brackets. Blockbootstrapped standard errors in curly brackets.

primogeniture as their succession rule during the period of our study, increasing its insti-tutionalization over time. Thus, regions that used primogeniture for more years are likelyto be those in which inheritance norms were better institutionalized. Column (3) of Table6 shows that the effect of percent male varies with the fraction of years that a polity usedprimogeniture in ways similar to how the effect varies with whether or not a polity usedSalic Law or Roman Law: the effect of percent male is weaker in polities that increased thedegree of institutionalization of their succession rule by adopting primogeniture.

The results of Table 6 columns (1) - (3) suggest that the degree of institutionalization ofsuccession rules matters more for explaining variation in the effect of percent male on con-temporary development, than does any possible variation in gender bias that might accountfor variation in the adoption of these norms. Indeed, political gender bias is likely to havebeen high across all of Europe. This view is supported by the fact that there is a total of only91 women rulers in our data (only 10.6% of the total number of rulers). It is also supportedby the result of column (4) in Table 6, which shows that the effect of percent male does notdiscernibly vary with the number of female rulers.

26

The key takeaway from the results of this section is that the effect of shocks and luck(reflected here in the availability of male heirs) are conditioned by the institutional environ-ment. Specifically, luck would play more of a role in weakly institutionalized environments.So a key function of institutions is to mitigate the impact of chance events. Western Europe’slong run economic success is, we argue, at least in some part due to the ability of its rulersand subjects to control the vagaries of fate by designing institutions that resolve coordinationproblems. In the next section, we show, however, that the degree of institutionalization mayitself be a consequence of chance events.

6.4 Endogenous Institutionalization of Succession Rules

Most European polities eventually adopted primogeniture as their succession norm at somepoint during the five-hundred-year period of our analysis (Kokkonen and Sundell, 2014; Blay-des and Chaney, 2013). However, according to Kokkonen and Sundell (2014), primogeniturewas often abandoned in many cases even after it was adopted. For example, according toKokkonen and Sundell (2014), Bohemia adopted primogeniture in 1230, abandoned it in1305, readopted it in 1346, and abandoned it in 1419. Notably, both adoptions of primo-geniture occurred under sovereigns with multiple male heirs, while both abandonments ofit occurred under monarchs with no children. The experience of Bohemia suggests thatthe availability of male heirs could be an important determinant of which regions adoptedprimogeniture early, or for longer periods.

We find indirect evidence for the hypothesis that polities might have strategically adoptedand abandoned primogeniture depending on the abundance or shortage of male heirs. Forexample, in a cross sectional OLS specification, percent male has a positive effect of onthe fraction of years in the five hundred year period of our analysis that a polity usedprimogeniture as its succession rule (coef. = 2.349, s.e. = 0.660, macro-polity clustereds.e. = 1.103, block bootstrapped s.e. = 1.176, N = 124, R2 = 0.094). In addition to this, wealso find in a panel regression with decade fixed effects that the adoption of primogeniture(1 if adopted, 0 if not) is positively affected by the legitimate number of male heirs (coef. =0.0022, s.e. = 0.0003, region clustered s.e. = 0.007, number of regions = 117, R2 = 0.99).These results provide evidence that the availability of male heirs also affected the speed anddegree of institutionalization of succession norms.

27

7 Conclusion

Europe today—prosperous as it is in comparison to many other parts of the world—exhibitsa great deal of internal variation in development levels across its regions. In this paper,we argued that a substantial part of this variation is due to the uneven development ofstate institutions across medieval European polities. We showed that the likelihood of theavailability of male heirs to a European region’s monarchies in the period 1000-1500AD hasa positive effect on contemporary development in that region. Since the unavailability ofmale heirs was an important driver of internal conflict and political instability, which in turnaffected state development, our approach enabled us to sidestep the problem that reliablemeasures of medieval state development are scarce, as well as the fact that medieval statedevelopment could have been affected by a host of unobservable region characteristics thatdirectly affect contemporary development.

Besides emphasizing the importance of state building in general, our results show the pre-1500 period specifically was an important period in the political development of the modernworld, and that within Europe the political trajectories of regions may have diverged muchearlier than is sometimes argued. The emergence of the first modern states in this period wasso important, and the states themselves so fragile, that even small disruptions could have longterm consequences—consequences that we have shown are measurable even after centuries ofrevolution, industrialization, war and institutional growth. In addition, our results reinforcethe findings of the political economy of development literature on the negative effects ofviolent conflict, and the importance of political stability for development.

Finally, our findings emphasize the importance of chance, and how chance works incombination with both culture and institutions in shaping development paths. In regionswhere accidents of male birth allowed for a series of uncontested leadership transitions, rulerswere able to build the state institutions necessary to support economic development. In areasburdened by a greater potential for politically instability, the path to economic prosperitywas much more arduous. Far from being determined solely by natural resources, diseaseenvironments, preexisting political institutions, or even the plans of their rulers, the fortunesof regions like Naples and France were influenced by accidents of biology. As such, theresults provide both a rejoinder to a focus on large structural predictors of social scientificphenomena, reminding us of the chaos of politics in an unpredictable world.

28

References

Abulafia, David. 1988. Frederick II: a medieval emperor. Oxford University Press, USA.

Acemoglu, Daron, Davide Cantoni, Simon Johnson and James A Robinson. 2011. “TheConsequences of Radical Reform: The French Revolution.” American Economic Review101(7):3286–3307.

Acemoglu, Daron, Simon Johnson and James A Robinson. 2002. “Reversal of fortune: Geog-raphy and institutions in the making of the modern world income distribution.” Quarterlyjournal of economics pp. 1231–1294.

Acemoglu, Daron, Simon Johnson and James A Robinson. 2005. “Institutions as a funda-mental cause of long-run growth.” Handbook of economic growth 1:385–472.

Alesina, Alberto, Paola Giuliano and Nathan Nunn. 2013. “On the Origins of Gender Roles:Women and the Plough.” The Quarterly Journal of Economics 128(2):469–530.

Alesina, Alberto and Roberto Perotti. 1996. “Income distribution, political instability, andinvestment.” European Economic Review 40(6):1203–1228.

Anderson, Perry. 1979. “Lineages of the absolutist state.”.

Angrist, Joshua D and Jörn-Steffen Pischke. 2008. Mostly harmless econometrics: An em-piricist’s companion. Princeton university press.

Astarita, Tommaso. 2002. The continuity of feudal power: the Caracciolo di Brienza inSpanish Naples. Cambridge University Press.

Austin, Gareth. 2008. “The ‘reversal of fortune’thesis and the compression of history: per-spectives from African and comparative economic history.” Journal of International De-velopment: The Journal of the Development Studies Association 20(8):996–1027.

Bairoch, Paul. 1991. Cities and economic development: from the dawn of history to thepresent. University of Chicago Press.

Baldwin, John W. 1991. The government of Philip Augustus: foundations of French royalpower in the middle ages. Univ of California Press.

Banfield, Edward C. 1967. “The moral basis of a backward society.”.

Barro, Robert J. 1996. “Democracy and growth.” Journal of economic growth 1(1):1–27.

29

Bates, Robert H. 2001. Prosperity and violence. WW Norton.

Beik, William. 1985. Absolutism and society in seventeenth-century France: State power andprovincial aristocracy in Languedoc. Cambridge University Press.

Besley, Timothy and Torsten Persson. 2010. “State capacity, conflict, and development.”Econometrica 78(1):1–34.

Besley, Timothy and Torsten Persson. 2011. Pillars of prosperity: The political economicsof development clusters. Princeton University Press.

Blaydes, Lisa and Eric Chaney. 2013. “The feudal revolution and europe’s rise: Politicaldivergence of the christian west and the muslim world before 1500 ce.” American PoliticalScience Review 107(01):16–34.

Bonney, Richard. 1995. Economic Systems and State Finance: The Origins of the ModernState in Europe 13th to 18th Centuries. Oxford University Press.

Boserup, Ester. 1970. Woman’s role in economic development. Earthscan.

Brewer, John. 1990. The sinews of power: War, money, and the English state, 1688-1783.Harvard University Press.

Chaney, Eric. 2014. “The Medieval Origins of Comparative European Development: Evidencefrom the Kingdom of Navarre.” Working Paper .

Charron, Nicholas, Lewis Dijkstra and Victor Lapuente. 2014. “Regional governance matters:quality of government within European Union member states.” Regional Studies 48(1):68–90.

Classen, Albrecht, David Ditchburn, Simon Maclean and Angus Mackay. 1997. Atlas ofmedieval Europe. Psychology Press.

Colletta, Nat J and Michelle L Cullen. 2000. Violent conflict and the transformation of socialcapital: Lessons from Cambodia, Rwanda, Guatemala, and Somalia. Vol. 795 World BankPublications.

Collier, Ruth Berins and David Collier. 2002. Shaping the political arena. University of NotreDame Press Notre Dame, IN.

Corcos, Christine Alice. 2012. “From Agnatic Succession to Absolute Primogeniture.” Michi-gan State Law Review p. 1587.

30

Cox, Gary W, Douglass C North and Barry R Weingast. 2015. “The violence trap: Apolitical-economic approach to the problems of development.” Working Paper.URL: http://papers.ssrn.com/sol3/papers.cfm?abstract-id=2370622

DeLong, J Bradford and Andrei Shleifer. 1993. “Princes and Merchants: City Growth Beforethe Industrial Revolution.” Journal of Law and Economics 36:671–702.

Dincecco, Mark and Massimiliano Onorato. 2013. “Military Conflict and the Economic Rise ofUrban Europe.” http://papers.ssrn.com/sol3/papers.cfm?abstract\_id=2341713.

Downing, Brian M. 1993. The military revolution and political change: Origins of democracyand autocracy in early modern Europe. Princeton University Press.

Dube, O and SP Harish. 2015. “Queens.” Working paper, New York University.URL: http://odube.net/wp-content/uploads/2015/11/Queens_Oct2015.pdf

Evans, Peter B. 1995. Embedded autonomy: states and industrial transformation. CambridgeUniv Press.

Grech, Victor, Charles Savona-Ventura and P Vassallo-Agius. 2002. “Unexplained differencesin sex ratios at birth in Europe and North America.” Bmj 324(7344):1010–1011.

Greif, Avner. 1993. “Contract enforceability and economic institutions in early trade: TheMaghribi traders’ coalition.” The American economic review pp. 525–548.

Guiso, Luigi, Paola Sapienza and Luigi Zingales. 2016. “Long-term persistence.” Journal ofthe European Economic Association 14(6):1401–1436.

Harding, Alan. 2002. Medieval Law and the Foundations of the State. Oxford UP.

Hartung, John. 1985. “Matrilineal inheritance: New theory and analysis.” Behavioral andBrain Sciences 8(04):661–670.

Herbst, Jeffrey. 2000. States and power in Africa: Comparative lessons in authority andcontrol. Princeton University Press.

Herlihy, David. 1962. “Land, family and women in continental Europe, 701-1200.” Traditiopp. 89–120.

Herz, John H. 1952. “The problem of successorship in dictatorial regimes; A study in com-parative law and institutions.” The journal of Politics 14(01):19–40.

31

Jansen, Sharon L. 2002. The Monstrous Regiment of Women: female rulers in early modernEurope. Macmillan.

Jones, Michael. 2000. The New Cambridge Medieval History, Volume 6, c. 1300-c 1415.Cambridge University Press.

Kohn, George Childs. 2013. Dictionary of wars. Routledge.

Kokkonen, Andrej and Anders Sundell. 2014. “Delivering Stability? Primogeniture andAutocratic Survival in European Monarchies.” American Political Science Rev. pp. 1–16.

Kurrild-Klitgaard, Peter. 2000. “The constitutional economics of autocratic succession.”Public Choice 103(1-2):63–84.

Lewis, Andrew W. 1981. Royal succession in Capetian France: studies on familial order andthe state. Harvard University Press Cambridge/London.

Lipset, Seymour Martin and Stein Rokkan. 1967. Party systems and voter alignments:Cross-national perspectives. Vol. 7 Free press.

Maddison, Angus. 2007. Contours of the world economy 1-2030 AD: Essays in macro-economic history. Oxford University Press.

McLaughlin, Megan. 1990. “The woman warrior: gender, warfare and society in medievalEurope.” Women’s Studies: An Interdisciplinary Journal 17(3-4):193–209.

McNaughton, Arnold. 1973. Book of Kings: A Royal Genealogy. Quadrangle/New YorkTimes Book Co.

Migdal, Joel S. 1988. Strong societies and weak states: state-society relations and statecapabilities in the Third World. Princeton University Press.

Milgrom, Paul R, Douglass C North and Barry R Weingast. 1990. “The role of institutionsin the revival of trade.” Economics & Politics 2(1):1–23.

Morby, John E. 1989. Dynasties of the World. Oxford University Press.

North, Douglass C. 1973. The rise of the western world: A new economic history. CUP.

North, Douglass C. 1990. Institutions, institutional change and economic performance. Cam-bridge university press.

32

North, Douglass C and Barry R Weingast. 1989. “Constitutions and commitment: theevolution of institutions governing public choice in seventeenth-century England.” Thejournal of economic history 49(04):803–832.

North, Douglass Cecil. 1981. Structure and change in economic history. Norton.

Nunn, Nathan and Leonard Wantchekon. 2011. “The Slave Trade and the Origins of Mistrustin Africa.” American Economic Review 101(7):3221–52.

Pierson, Paul. 2000. “Increasing returns, path dependence, and the study of politics.” Amer-ican political science review 94(02):251–267.

Pomeranz, Kenneth. 2009. The great divergence: China, Europe, and the making of themodern world economy. Princeton University Press.

Potter, John Milton. 1937. “The development and significance of the Salic Law of the French.”English Historical Review pp. 235–253.

Putnam, Robert D, Robert Leonardi and Raffaella Y Nanetti. 1994. Making democracywork: Civic traditions in modern Italy. Princeton university press.

Schechtman, Joseph B. 1953. “Postwar population transfers in Europe: a survey.” The Reviewof Politics 15(02):151–178.

Skocpol, Theda. 1979. States and social revolutions: A comparative analysis of France,Russia and China. Cambridge University Press.

Stasavage, David. 2014. “Was Weber Right? The Role of Urban Autonomy in Europe’sRise.” American Political Science Review 108(02):337–354.

Strayer, Joseph R. 1970. On the medieval origins of the modern state. Princeton UP.

Sumption, Jonathan. 2009. The Hundred Years War. Vol. 3 U of Pennsylvania Press.

Tabellini, Guido. 2010. “Culture and institutions: economic development in the regions ofEurope.” Journal of the European Economic Association 8(4):677–716.

Takayama, Hiroshi. 1993. The administration of the Norman kingdom of Sicily. Vol. 3 Brill.

Thelen, Kathleen. 2004. How institutions evolve. Cambridge UP.

Tilly, Charles. 1992. Coercion, capital, and European states, AD 990-1992. Blackwell Oxford.

Tullock, Gordon. 1987. Autocracy. Springer Science & Business Media.

33

Voigtländer, Nico and Hans-Joachim Voth. 2013. “The three horsemen of riches: Plague, war,and urbanization in early modern Europe.” Review of Economic Studies 80(2):774–811.

Ward, Jennifer. 2014. Women in Medieval Europe: 1200-1500. Routledge.

34